estimator of effective/relative wind velocity

Wind & Water Controls
1

Hi, all.
I am working on the controller for offshore floating wind turbines in Region 2 (between cut-in and rated velocity), and using FAST in Simulink.

I would like to have an estimator for obtaining the effective/relative wind velocity (considering platform’s moving…) at hub height.
Therefore, I chose the FAST’s output “RotTorq”, “RotSpeed” and the “Cq-lameda curve” to solve the equation as follows by iteration.

Tr=0.5phopiR^3Cq*Vrel^2

However, the result was questionable.

Then, I tried it on an onshore wind turbine, and prepared a wind file with a step from 8m/s to 6m/s. The result was still slow, and the estimator needed a transient process to move to 6m/s, about 20s.

Later, I tried it again on the onshore wind turbine using the same wind file above mentioned. The only difference is that I CLOSE the GenDOF. The result seemed good, and the estimator would suddenly move to 6m/s. But I still needed the GenDOF.

Could somebody tell me the reason? In my opinion, I use the torque and speed of the rotor, not the generator, so I cannot understand the result.

And could somebody tell me the method for obtaining the effective/relative wind speed? It is important to me.

Thanks in advance.

Best regards,
Lei DUAN

2

Hi,

When you say you closed the generator DOF, do you mean you switched it to False? This would mean then the rotor speed is now fixed. If the rotor speed is not fixed, then changes in wind speed will cause transients while the rotor speed accelerates/decelerates to a new state, which I think is the transient you’re seeing when you don’t fix rotor speed.

That said, I do think it should be possible to develop a better estimator even with turbine dynamics, but I think it will be important to include turbine acceleration. See for example the discussion of the issue it the introduction of this paper:

iopscience.iop.org/1742-6596/75/1/012082

Paul

3

Hi to all,

for the purposes of my research project I am trying to develop an effective wind speed estimator in Simulink for the 5 MW Fast Wind Turbine, in line with the work iopscience.iop.org/1742-6596/75/1/012082 appearing in the above post. My progress so far, is summarized as follows:

  • Assuming rigid drive-train (DrTrDOF=False in ElastoDyn ), the Aerodynamic torque Ta (Qa in the paper above) is computed in Simulink through a Matlab function block by the following relation:

                      Ta=Jr*omegar_dot+LSShftTq,

where Jr denotes the rotor inertia (I took Jr=115.926 kg*m^2, hub inertia about low speed shaft found in “Definition of a 5-MW Reference Wind Turbine for Offshore System Development”), omegar_dot is the acceleration of the rotor (I take RotSpeed as output from FAST convert it to rad/sec by multiplying with pi/30 and then taking the derivative using a derivative block in Simulink), and LSShftTq is given, again, as output from FAST.

  • Then from
    2Ta/rhopiR^5omegar^2=cp(beta,TSR)/TSR^3=F(beta,TSR)

if one computes Ta and omegar (aerodynamic torque and rotor speed), the left-side term of the quation above is known and F (beta,TSR) is computed.
The table F vs TSR can be computed offline, using the knowledge of Cp(beta,TSR), thus for a given F, the tip-spped ratio TSR can be computed.To compute F=Cp/TSR^3 i have used the tables found in "NREL 5-MW reference turbine - CP, CQ, CT Coefficients - #13 by Sebastian.Hippel My results for Cp and F, for beta=0 when the wind turbine operates below rated power, are attached.

  • Then, with TSR known the effective wind speed is given from V=omegar*R/TSR. I have developed all the previous steps in Simulink using a stepwise changing simulated wind speed, with wind speed below the rated wind speed. My concern is that, although the estimated wind speed converges to the simulated one there is a significant delay for the actual convergence which I can not explain. I have attached the comparison between the simulated wind speed and the estimated one. The actual response of the estimated wind speed resembles a lot the actual behavior of a low-pass filter- could dynamic inflow explain this behavior? Do you see any mistake i have done in my modelling design? I have also attached the AeroDyn file I have used for the simulations.

I am looking forward to your responses, thank you in advance for your comments.

Bart

Aerodyn.JPG
Aerodyn.JPG916×591 130 KB

Cp and F.JPG
Cp and F.JPG757×288 16.6 KB

Simulated wind speed vs estimated wind speed.jpg
Simulated wind speed vs estimated wind speed.jpg1588×782 30.8 KB

4

Dear Bert,

Regarding your calculation of Ta:

  • The hub inertia of the NREL 5-MW turbine is 115926 kgm^2, not 115.926 kgm^2. However, the equation you are using requires the rotor inertia instead of the hub inertia, which for the NREL 5-MW turbine is 38677040.613 kg*m^2.
  • When using AeroDyn v15, you can get the aerodynamic applied torque directly as an output i.e. RotAeroMxh.

Regarding the delay in the estimate:

  • Your AeroDyn model does not have any dynamics enabled, so, the delay is not coming from that.
  • Instead, the delay is likely caused by structural dynamics i.e. the structural degrees-of-freedom that are enabled.

I hope that helps.

Best regards,